Original Paper

Measurement and Control

2019, Vol. 52(7-8) 740–757
Effect of earthquake excitation on Ó The Author(s) 2019
Article reuse guidelines:
circular tunnels: Numerical and sagepub.com/journals-permissions
DOI: 10.1177/0020294019847705

experimental study journals.sagepub.com/home/mac

DK Singh1 , SR Karumanchi1, A Mandal1, YB Katpatal1 and A Usmani2

Abstract
This paper studies the behaviour of circular tunnel subjected to dynamic excitation. Tunnels with three different dia-
meters were selected to perform the shake table test at three different covers. The dry sandy soil was used for testing.
The mechanical properties like Young’s modulus and shear modulus of sand was calculated from bender element test.
The soil–tunnel interface coefficient was calculated from the direct shear test. The soil pressure generated due to
dynamic loading were measured by soil pressure transducers. The actual motion of shake table was captured by hand-
held vibration analyser. The tunnel was placed parallel and perpendicular to the direction of shaking. The three-
dimensional finite-element model was developed for tunnel with both the orientations. The tunnel was assumed to be
elastic. Dry sand was assumed to follow non-linear elasto-plastic material using Mohr–Coulomb failure criterion with
non-associated flow rule. The results obtained from numerical analysis are compared with experimental results and are
expressed in the form of peak dynamic stresses. The time history and fast Fourier transform results of dynamic stresses
are also compared. It shows reasonable agreement with both values. Finally, the seismic design guidelines for tunnel are
suggested.

Keywords
Shake table, tunnel, finite-element method, absorbing boundary, bender element

Date received: 23 March 2018; accepted: 18 March 2019

Introduction among researchers. Almost seven tunnels experienced

An earthquake occurs due to sudden release of energy,
resulting in seismic wave generation. These waves cause
large damage to infrastructure when it reaches to the
ground surface. Circular tunnels are often constructed
for highways, railways, water conveyance in hydro
power projects and metro projects. These tunnels are
mainly excavated using tunnel boring machine or drill
and blast methods. Earlier, theses tunnels were con-
structed in sound rock with high cover. However, due
to rapid urbanization and urgent need of society, there
is an increase in the demand for underground struc-
tures like tunnels. In addition, tunnels are now being
constructed in high seismic zones with soft ground con-
dition, mainly in soil. The safety of the tunnel is very
important during adverse loading conditions. This
demands more focused research to understand the
behaviour of tunnels subjected to seismic loading.
Earlier, underground structures were considered safe
during the earthquake. However, failure of under-
ground structures during the Kobe earthquake (1995,
Japan), Chi-Chi earthquake (1999 Taiwan) and
Kocaeli earthquake (1999 Turkey) changed the belief
serious damage during Kobe earthquake, 93 tunnels
were affected during the 1923 Kanto earthquake, and
Niigataken-Chuetsu earthquake affected 24 tunnels.1
The damage occurred either due to large deformation
or due to liquefaction of surrounding soil. Poor geolo-
gical conditions, shallow depth of tunnels and the pres-
ence of active faults were the main reason for the
damage of underground structures. While seismic exci-
tation of tunnel induces axial and curvature deforma-
tion in longitudinal direction, it induces ovaling or
raking in transverse direction, depending on the shape
of the tunnel. Axial and curvature deformations were
reported in tunnels placed along the longitudinal direc-
tion of shake table shaking.2 Moreover, lack of suitable
1
Department of Civil Engineering, Visvesvaraya National Institute of
Technology, Nagpur, India
2
Engineers India Limited, New Delhi, India
Corresponding author:
DK Singh, Department of Civil Engineering, Visvesvaraya National
Institute of Technology, Nagpur 440010, India.
Email: dhananjay.singh86@gmail.com

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
Singh et al. 741

guidance in most of the cases lead to inefficient design shear test. Furthermore, the numerical model was
by professionals. Full dynamic analysis should be done established in the finite-element model. The results
to know the effect of soil–structure interaction. obtained from numerical analysis was compared with
Simplified dynamic analyses should be used for prelimi- experimental results and are expressed in the form of
nary design. percentage change in peak dynamics stresses with
In the field, modern instruments like load cell and respect to static pressure on the tunnel. The time history
pressure cell provide useful data and insight into the and fast Fourier transform (FFT) results of dynamic
behaviour of underground structures. It is compara- stresses are also compared.
tively convenient to use the instruments and sensors
installed on site to get the load and pressure data of
tunnel in case of static analysis. However, it is not as
Experimental setup
convenient to measure dynamic forces at the site Shake table
because high magnitude earthquakes are rare and
The shake table used for testing is unidirectional with
vibration from other sources also affects the dynamic
table top dimension of 1.5 m 3 1.5 m (Figure 1). The
force measurement significantly. Sensors for dynamic
motion of table is constrained to one-dimension only.
measurement should be carefully selected to avoid mea-
The lateral displacement of shake table is given through
surement issues.3 Care should be taken to select the
an electric motor of 15-HP capacity. The maximum dis-
scale of an experimental model to predict the behaviour
placement of shake table is 650 mm and maximum fre-
of prototype.4 To overcome this limitation of field
quency is 10 Hz under ideal conditions. The maximum
data, lab tests were performed on models.
load-carrying capacity of shake table is 2000 kg.
Experimental and numerical simulation of the tunnel
However, increasing the load or displacement of shake
has been investigated by many researchers.5–12 A series
table decreases the frequency. Therefore, for the pres-
of shake table test of the utility tunnel in transverse and
ent test, the displacement was set to 65, 610, 615,
longitudinal direction was performed; the results show
620, 625 and 650 mm, whereas the frequency was set
that the amplification factor of soil decreases with
0.5, 0.8, 1.5 and 2.0 Hz. It should be noted that 2.0 Hz
increase in peak ground acceleration (PGA) due to soil
frequency is available for only 65 mm displacement,
non-linearity,7 and high-intensity shaking generates
whereas for 650 mm displacement 0.5 and 0.8 Hz fre-
some residual stress in the tunnel.10 The acceleration
quencies are only available.
response of tunnel is higher than surrounding soil for
high shaking.6 Under seismic loading, the tunnel
vibrates in rocking mode, and racking13 also bending Soil container/tank
deformation occurs in the tunnel.6 However, most of
The rigid type container was used for experiment. The
the researchers are more focused on internal forces,
internal dimensions are 1.0 m 3 1.0 m 3 1.0 m, and it is
deformation and acceleration developed in tunnel dur-
firmly fixed to the shake table. It was fabricated using
ing seismic activity, very few studies have been carried
on the stresses generated in the soil during earthquake.
Since, the value of soil pressure exerted on tunnel is
used by the designer for support design of tunnels. This
study is more focused on the soil pressure generated
during the earthquake.
Frictional characteristics of the interface between
soil and tunnel significantly affect load transfer
mechanism between soil and tunnel. To design tunnel
for seismic loading, closed form solutions are available
for full slip and no slip interface condition, and the
designed tunnel is to be checked for both the cases.
However, partially slip interface condition may gener-
ate higher forces in tunnel.14 Therefore, the closed form
solution cannot predict the actual behaviour of tunnel
during seismic loading, and full dynamic analysis
should be performed.15
In this paper, the 1g shake table test was performed
on the circular tunnel subjected to dynamic loading.
The test was performed with different tunnel diameters
at different covers. The dynamic pressure developed
due to seismic force in dry sand is measured by soil
pressure transducers. In addition, the actual motion of
shake table is captured by hand-held vibration analyser.
The soil–tunnel interface value was calculated by direct Figure 1. Shake table with container.
742 Measurement and Control 52(7-8)

Figure 2. Experimental setup and arrangement of sensors for transverse direction of tunnel.

Figure 3. Arrangement of sensors for longitudinal orientation of tunnel: (a) experimental and (b) graphic diagram.

perspex glass of 18 mm thickness and strengthened Methodology

using flat and angle structural steel sections (Figure 1).
The expanded polyethylene (EPE) foam was used on
the inner boundary of the container, perpendicular to
the direction of shake table movement. The base was
made rough by sticking sand using industrial adhesives.
Sensors setup
Two types of sensors were used to find the seismic
response of the tunnel. The hand-held vibration sensors
with magnetic tip were placed at the centre of the verti-
cal boundary of the container/tank (Figure 2). The
hand-held vibration analyser was used to capture the
actual motion of the shake table.
Three soil pressure transducers were used to measure
the pressure in the soil and on tunnel. The data were
recorded on the dynamic data acquisition module. The
arrangements of sensors are shown in Figures 2 and 3.
Soil filling in container/tank
The dry sandy soil was used as fill material for a tunnel
in the shake table experiments. Permanent changes can
appear in internal forces because of densification of
sand layer during intense shaking.5 Therefore, prior to
start of the actual test, the sand is compacted to its
maximum possible density, to avoid sand densification
during the test. To achieve this, the sand was placed
loosely by air pluviation, then the vibration was given
from shake table. Three layers of sand were chosen for
simplification in the analysis. Total five such trials were
made. It was observed that during shaking, the sand
densification occurs (12%–18%). After sand densifica-
tion, the densities at the middle of each layer were
measured and found to be 1.55, 1.65 and 1.75 g/cm3
(Figure 4). The overall depth of the sand layer is 85 cm.
To achieve the desired density, the soil in the container
was divided into three layers: top and middle layer
Singh et al.
Figure 4. Dry sand layers.
having a thickness 28 cm and bottom layer thickness as
29 cm. The target was set to maintain the sand density
at 1.55, 1.65 and 1.75 g/cm3 for top, middle and bottom
layers, respectively. Additional tamping was done to
achieve the desired density. After this arrangement, the
sand densification during the test nearly ends.
Therefore, for numerical modelling, the sand layer is
divided into three parts, and the density in the middle
of each layer is considered for the analysis (Figure 4).
The tunnel was made up of perspex glass having an
outer diameter of 10, 15 and 20 cm with thickness of 3,
3 and 6 mm, respectively. The shake table experiments
were conducted by placing the tunnels in parallel and
perpendicular (Figures 2 and 3) to the direction of shak-
ing (one at a time) at three different covers.
Absorbing boundary
The effect of artificial boundaries of a soil container on
dynamic response of soil can be significant if not
designed properly. Use of absorbing material on
boundary is recommended for minimizing the bound-
ary effect. The commercially available EPE foam panel
was used as the absorbing boundary in the present
tests. These foams were placed on both inner sides of
the end walls of the soil container, perpendicular to the
shaking direction. The thickness of foam was designed
as per practical guidelines provided by Lombardi; for
this case, it is 25mm. The design criteria used here are
based on impedance. For good absorbing boundary,
the impedance of soil should be greater than 200 times
the impedance of foam.16
To determine the damping associated with foam, the
hysteresis loss of energy in the foam was calculated as
per ASTM D 3574-17 (Figure 5). The compression
force displacement (CFD) procedure was followed.17
The hysteresis loss calculated as per equation (1) is
25%
A1
Hysteresis loss = A1 + A2 ð1Þ
where A1 is the area in between loading–unloading
curve, and A2 is the area under unloading curve.
743
Figure 5. Hysteresis loss in foam.
Material properties
The bender element test was done to get the mechanical
properties like Young’s modulus, shear modulus and
Poisson’s ratio of sand and foam. Figure 6 shows the
test results obtained from bender element. In this test,
the p-wave and s-wave velocity is computed by dividing
the length of sample with time taken by waves reach
from the source end to receiver end. Time taken to
reach from source to receiver was the difference
between the arrival times of the first peak at receiver
end to the peak at source end. For present test, the
sample length was 12.0 cm and diameter 6.0 cm. These
dimensions are selected to obtain the best results from
bender element tests as a slenderness ratio greater than
2 gives best results.18 However, bender element source
and receiver ends have the outward projection of
1.5 mm each. Therefore, tip-to-tip distance between
source and receiver is 11.7 cm. The test was conducted
for different densities of sand like 1.55, 1.65 and 1.75 g/
cm3. The dilatation angle of dry sand was assumed to
be 1°. In addition, the value of D60, D30 and D10 was
0.85, 0.65 and 0.45 mm, respectively.
The bender element can be used for materials treated
with air foam, that is, the material having greater quan-
tity of air voids.19 The test similar to the sand was per-
formed on industrial foam which was used as absorbing
boundary. Here, the tip-to-tip distance between source
and receiver is 6.2 cm. For industrial foam, the s-wave
is showing the peak clearly. However, the p-wave
obtained during the test does not show any peak.
Therefore, the method of cross-correlation was used on
p-wave to find the p-wave velocity (Figure 6(d)).
Based on the p-wave and s-wave velocity obtained
from bender element test, the shear modulus, Poisson’s
ratio and Young’s modulus of the materials is calcu-
lated using equations (2)–(4), respectively
G = V2s r ð2Þ
Vp 2
ð Þ
Vs 2
n= Vp 2
ð3Þ
2ðÞ
Vs 1
744 Measurement and Control 52(7-8)

Figure 6. Bender element test results: (a) sand 1.55 g/cm3, (b) sand 1.65 g/cm3, (c) sand 1.75 g/cm3 and (d) foam.

Table 1. Material properties.

Density Elastic modulus Poisson’s ratio Inner friction Cohesion Specific emax emin
(kg/m3) (kPa) angle (°) (kPa) gravity

Sand (top layer) 1550 27,453 0.29 30 0 2.69 0.79 0.49

Sand (middle layer) 1650 145,570 0.28 34 0
Sand (bottom layer) 1750 202,642 0.27 37 0
Tunnel 1190 3,102,640 0.37 – – 1.19 – –
Foam 22 2396 0.25 – – – – –

Figure 7. Direct shear test setup.

Singh et al.
V2p rð1 + nÞð12nÞ
ð4Þ
E= 1n
where G is the shear modulus, r is the density of the
material, Vs is s-wave velocity, n is Poisson’s ratio, Vp
is p-wave velocity and E is Young’s modulus.
Tensile test was conducted on perspex glass rod to
calculate Young’s modulus of the tunnel and container.
Table 1 shows the properties of various materials used
in the experiment.
The value of soil–tunnel interaction coefficient is cal-
culated by performing the direct shear test. The dimen-
sion of the direct shear box is 6 cm 3 6 cm with a height
Figure 8. Continued.
745
of 2.4 cm. The perspex glass plate of thickness 12 mm
was placed on the lower side of the direct shear box,
and the upper part is filled with the sand with 90% of
relative density (Figure 7). The soil–tunnel interface
coefficient of friction was calculated to be 0.12.
Input motion
The hand-held vibration analyser was used to capture
the actual motion of the shake table. The maximum
frequency of 100 Hz was set to capture with interval of
0.25 Hz. The final data shown by the hand-held
746 Measurement and Control 52(7-8)

Figure 8. Frequency spectrum of shake table motion: (a) 5 mm – 0.5 Hz, (b) 5 mm – 0.8 Hz, (c) 5 mm – 1.5 Hz, (d) 5 mm – 2.0 Hz,
(e) 10 mm – 0.5 Hz, (f) 10 mm – 0.8 Hz, (g) 10 mm – 1.5 Hz, (h) 15 mm – 0.5 Hz, (i) 15 mm – 0.8 Hz, (j) 15 mm – 1.5 Hz, (k) 20 mm –
0.5 Hz, (l) 20 mm – 0.8 Hz, (m) 20 mm – 1.5 Hz, (n) 25 mm – 0.5 Hz, (o) 25 mm – 0.8 Hz, (p) 25 mm – 1.5 Hz, (q) 50 mm – 0.5 Hz and
(r) 50 mm – 0.8 Hz.
Singh et al.
Figure 9. Conversion of shake table frequency domain motion
into time domain motion for 50-mm displacement and 0.5-Hz
frequency: (a) frequency domain and (b) time domain.
vibration analyser are the average of data obtained by
10 cycles of shake table movement. It is assumed that
the vibration captured by the hand-held vibration ana-
lyser is purely from the shake table, the cut-off fre-
quency of input motion was set to 25 Hz for numerical
analysis. The test is performed by changing the shake
table displacement and its frequencies. Figure 8 shows
the frequency spectrum for various motions of shake
table during the test.
The hand-held vibration analyser captured the fre-
quency and amplitude of shake table motion. Non-lin-
ear dynamic analysis is complex in the frequency
domain. Therefore, to convert the data from frequency
domain to time domain, a MATLAB code was devel-
oped. Figure 9 shows a sample for conversion from fre-
quency domain to time domain for 50-mm
displacement with 0.5 Hz frequency. In addition, the
shake table motion closely simulates the actual earth-
quake in terms of predominant frequency. For actual
earthquake, predominant frequency is near 1.5 Hz and
the frequency range of shake table tests are from 0.5 to
1.5 Hz.
Experimental test
The soil container was firmly fixed to the shake table.
The dry sandy soil was placed in container in three
layers by air pluviation. Additional tamping was given
to achieve the desired density (Figure 4). The tunnels
were placed in the soil at the cover of 10, 30 and 50 cm,
in parallel and perpendicular to the direction of shake
table motion. The sensors were arranged as shown in
747
Figures 2 and 3. The seismic force was induced by
changing the displacement and frequencies as explained
earlier. The dynamic soil pressure generated during
each motion was captured. There were in total, 18 sets
of tests for single arrangement of the tunnel and there
were total 18 such arrangements (2Orientation 3 3Cover 3
3Different Diameter Tunnels = 18).
Numerical modelling
The three-dimensional (3D) numerical model was pre-
pared for tunnels placed in the transverse direction and
longitudinal direction. The numerical model was made
in finite-element software ABAQUS. This software is
suitable and verified for use in static and dynamic anal-
ysis of underground structures.14 In 3D model, the tun-
nel is made with four-node shell element. Soil, foam
and container are modelled with eight-node brick ele-
ments of minimum element size 0.01 m and maximum
element size 0.1 m (Figure 8). The soil–tunnel interac-
tion was modelled as coulomb friction. The slippage
between soil and foam is ignored. The master–slave sur-
faces are used to simulate the soil–tunnel interaction
with coefficient of friction as 0.12. The input motion
was applied at the base of the container. The frequency
dependent Rayleigh damping factors a and b were cal-
culated by adopting the damping ratio of 5% for soil
and 25% for foam. The predominant frequencies were
taken by performing linear dynamic analysis on the 3D
model of the tunnel.
The base motion used for numerical model is the one
which was captured by the hand-held vibration analyser
in frequency domain and later converted to the time
domain (Figure 10). The Mohr–Coulomb material
model was used for dry sand in the present analysis,
and the tunnel was modelled as elastic material.
To perform an analysis close to the real state, the
geostatic stress is calculated first. In this stage, the ele-
ments of the tunnel are deactivated, this produces the
vertical displacement in order of 10–6 m. In the second
stage, the soil is excavated by deactivating the elements
in the region of excavation, simultaneously the elements
of the tunnel are activated, providing the support to the
excavated soil. In the third stage, the earthquake
dynamic load is applied at the base of the model. Time
discretization for incremental calculation is small
enough to achieve a stable and accurate solution; for
present analysis, it is considered as 0.001 s. The cohe-
sion of 1 kPa was used for sand to avoid numerical
instability problems.
Results and discussion
Effectiveness of absorbing boundary
The efficiency of absorbing boundary was checked by
measuring the stresses at two different elevations (at 10
and 50 cm from top) in soil without tunnel. The overall
depth of sand layer was 85 cm. Three soil pressures were
748 Measurement and Control 52(7-8)

Figure 10. 3D model of tunnel placed in (a) transverse direction and (b) longitudinal direction.

Table 2. Effectiveness of absorbing boundary by comparing the

horizontal stress at different location.

Sr. no. Overburden Stress (kPa) at different

depth distances from absorbing boundary
0 cm 24 cm 48 cm

1 10 cm 4.23 4.05 4.10

2 50 cm 2.12 2.01 2.04

installed at different distances from absorbing bound-

ary (at 0, 24 and 48 cm). Table 2 shows the results of
horizontal stress in direction of shaking (25 mm displa-
cement with 1.5 Hz frequency), recorded to check the
efficiency of absorbing boundary. There is 65% devia-
tion in peak dynamic stress value, which is acceptable. Figure 11. Acceleration response of soil container for shake
table displacement of 25 mm and frequency 1.5 Hz.
All other shaking intensities are smaller than 25-mm
displacement and 1.5-Hz frequencies, so deviation in
peak dynamic stress is even smaller. For shallow cover,
the horizontal dynamic stress is more than larger cover. Table 3. Flexibility ratio of tunnels in different soil layers with
different diameters.

Soil layers (density) Flexibility ratio

Rigidity of soil container
10 cm 15 cm 20 cm
In the experimental work and numerical analysis, the
soil container was assumed to be rigid. For present Top layer (1.55 g/cm3) 55 185 55
numerical simulation acceleration response of soil con- Middle layer (1.65 g/cm3) 593 989 593
Bottom layer (1.75 g/cm3) 411 1387 411
tainer was checked at two locations (Figure 11). One
location where hand-held vibration analyser is attached
(Figure 2) and other location is directly opposite to it.
In addition, the data recorded from hand-held vibra- wall) and from experimental work is matched. This
tion analyser is shown in Figure 11. The acceleration shows that the soil container is rigid as the phase lag of
response of numerical simulation for soil containers at acceleration response between two vertical walls is
two locations (middle section of right and left vertical negligible.
Singh et al.
Flexibility ratio of tunnel
To determine the relative stiffness between a circular
tunnel lining and the soil medium, the flexibility ratio is
defined by equation (5). Flexibility ratio measures the
ability of the tunnel to resist the distortion imposed
from soil medium during seismic load.
Es ð1n2t ÞR3
Flexibility ratio = ð5Þ
6Et Ið1 + nm Þ
where Es is the modulus of elasticity of the medium, nm
is Poisson’s ratio of the medium, Et is the modulus of
Figure 12. Continued.
749
elasticity of the tunnel lining, nt is Poisson’s ratio of the
tunnel, R is the radius of the tunnel lining, t is the thick-
ness of the tunnel lining and I is the moment of inertia
of the tunnel lining (per unit width). Table 3 shows the
flexibility ratio of three different diameters of the tun-
nel. The flexibility ratio is calculated for tunnel placed
in three different layers. Wide variety of flexibility ratio
of tunnel was used for present test. However, the flexi-
bility ratio shown is valid for tunnel embedded in single
layer. Tunnel embedded in two layers may have differ-
ent flexibility ratio.
750 Measurement and Control 52(7-8)

Figure 12. Peak dynamic stresses in fine sandy soil for tunnel in longitudinal direction: (a) T10/10, (b) T10/30, (c) T10/50, (d) T15/10, (e)
T15/30, (f) T15/50, (g) T20/10, (h) T20/30 and (i) T20/50.

Peak dynamic stresses from experiment and

numerical model
The results from the shake table experiment and numer-
ical modelling are represented in the form of graphs.
The stresses are expressed in percentage increase of
peak dynamic stresses with respect to the corresponding
value of stress generated during static analysis. The per-
centage change in peak dynamic stresses is plotted
against the input motion of shake table (displacement
(mm) – frequency (Hz)). Some discrepancies may come
due to recording issues that are present in soil pressure
transducers.13 In addition, the relative stiffness of sen-
sing plate and effect of grain size distribution may affect
the results.20 In general, peak dynamic stresses from
numerical analysis are in good agreement with the
experimental results. However, slight difference in Figure 13. Average increase in peak dynamic stresses with
results from experiments and numerical model is respect to static pressure versus C/D ratio for tunnel in
observed for 10 cm cover. longitudinal direction.

Tunnel in longitudinal direction. Figure 12 shows the per- generates less dynamic stresses. Increase in cover-to-
centage increase in peak dynamic stresses for different diameter (C/D) ratio decreases the dynamic stresses.
diameters of tunnel at different cover. The tunnel is Large diameter tunnels are subjected to less dynamic
placed in longitudinal direction. Flexible tunnel stresses with the same flexibility ratio at same cover.
Singh et al.
Figure 13 shows the average increase in peak
dynamic stresses with respect to static pressure versus
C/D ratio for tunnel in longitudinal direction. It was
observed that vertical dynamic stresses are predomi-
nant. For C/D \ 3.0 S1 is more than S3. Tunnel in
longitudinal direction shows, rocking behaviour of the
tunnel, and it increases with an increase in intensity of
shaking.
Tunnel in transverse direction. Figure 14 shows the percent-
age increase in peak dynamic stresses for different dia-
meters of tunnel at different cover. The tunnel is placed
in transverse direction. The abscissa represents various
Figure 14. Continued.
751
loading conditions, whereas ordinate shows the normal-
ized peak dynamic stresses value. It was observed that
tunnel with same flexibility ratios (T10 and T20), smaller
diameter tunnel generates more stresses, about three to
four times higher. With the increase in C/D ratio, the
dynamic stresses decrease. In addition, increase in flexi-
bility ratio decreases the peak dynamic stress. T10 shows
significant increase in vertical dynamic stresses at a fre-
quency of 1.5 Hz.
Figure 15 shows the average increase in peak
dynamic stresses with respect to static pressure versus
C/D ratio for tunnel in transverse direction. Vertical
stresses are predominant for smaller diameter of the
752
Figure 14. Peak dynamic stresses in fine sandy soil for tunnel in transverse direction: (a) T10/10, (b) T10/30, (c) T10/50, (d) T15/10, (e)
T15/30, (f) T15/50, (g) T20/10, (h) T20/30 and (i) T20/50.
Figure 15. Average increase in peak dynamic stresses with
respect to static pressure versus C/D ratio for tunnel in
transverse direction.
tunnel (T10) and at shallow depth. In addition, with an
increase in depth, vertical stresses are becoming predo-
minant. For less flexible tunnel (T10 and T20) S1 is
greater than S3.
Measurement and Control 52(7-8)
Soil pressure transducer response. Figures 16–19 shows the
plots of dynamic stresses from experiment and numeri-
cal analysis for 15cm tunnel diameter, 30 cm cover and
input motion of 625 mm displacement and 1.5 Hz fre-
quency. This is the strongest input motion generated
from the shake table for present test and this is the rea-
son for its selection. The time history results are com-
pared for tunnel in longitudinal (Figure 16) and
transverse (Figure 17) direction. In addition, the results
are compared in terms of FFT for tunnel in longitudi-
nal (Figure 18) and transverse (Figure 19) direction.
The comparison shows that numerical predictions are
in the same order of magnitude with the experimental
results. The difference recognized is due to the small
settlement of 3 mm occurs during the dynamic analysis
step in numerical modelling. However, the proposed
model captures peak dynamic stresses well. And the
design of civil engineering structures is based on the
peak stresses.
Tunnel design guidelines. Most codes follow a dual-design
philosophy for earthquake-resistant design of build-
ings. First is a design-based earthquake (DBE), which
is expected to occur during the design life of the
Singh et al. 753

Figure 16. Time versus dynamic stresses for tunnel in longitudinal direction having diameter 15 cm and overburden depth as
30 cm: (a) Sensor S1, (b) Sensor S2 and (c) Sensor S3.

Figure 17. Time versus dynamic stresses for tunnel transverse direction having diameter 15 cm and overburden depth as 30 cm: (a)
Sensor S1, (b) Sensor S2 and (c) Sensor S3.
754 Measurement and Control 52(7-8)

Figure 18. Frequency versus dynamic stresses for tunnel longitudinal direction having diameter 15 cm and overburden depth as
30 cm: (a) Sensor S1, (b) Sensor S2 and (c) Sensor S3.

building. In such earthquakes, the structure should get response of tunnel greatly depends on the surrounding
only minor or moderate damage. Second is maximum soil, C/D ratio and the flexibility ratio of tunnel.
considered earthquake (MCE), which is expected that In the present shake table tests, total 18 different
the structural damage should not result in total ground motions were used with a wide variety of PGA
collapse. motion (0.01g–1.20g). Zone-II and zone-V have six
According IS 1893:2016, India is divided into four motions each, whereas zone-III and zone-IV have three
zones (zone-II, -III, -IV and -V) of the earthquake. motions each (Figure 20).
Zone-II is associated with the lowest level of seismicity,
whereas Zone-V expects the highest level of seismi-
city.21 The zone factors used in the design of civil engi- Design steps
neering is 0.10g, 0.16g, 0.24g and 0.36g for zone-II, - 1. Get the seismic zone factor as per IS 1893: 2016
III, -IV and –V, respectively. These zone factors are (Table 4).
based on PGA of an MCE. Furthermore, these zones 2. Select the stress amplification coefficient
are created based on Medvedev–Sponheuer–Karnik (Table 5).
(MSK) intensity scale, where zone-V corresponds to 3. Use the equation (6) to calculate the maximum
intensity above IX. Rupture in underground pipeline dynamic soil pressure
can be observed at this intensity. However, there are no
z
special provisions for design of underground structures Sdyn(MCE) = K Sstatic ð6Þ
in Indian code as well as in national codes of other
countries. where Sdyn(MCE) is the peak dynamic pressure for MCE,
IS 4880:Part-5 states that if seismic force is signifi- z is the zone factor, K is the stress amplification coeffi-
cant, it should be considered in the design. In addition, cient, and Sstatic is the static pressure on the tunnel.
for extreme loading condition, the stresses on the tun-
nel lining from the soil is to be increased by 33.3%.22 4. Use the equation (7) to calculate the design-
Seismic force should be considered for the underground based maximum dynamic soil pressure
utilities tunnel if design earthquake spectral response
acceleration of a site is more than 0.33.23 However, the SdynðDBEÞ = 23 SdynðMCEÞ ð7Þ
Singh et al.
Figure 19. Frequency versus dynamic stresses for tunnel transverse direction having diameter 15 cm and overburden depth as
30 cm: (a) Sensor S1, (b) Sensor S2 and (c) Sensor S3.
Figure 20. Input motion versus peak ground acceleration
(zones are as per Indian Standard 1893:2016).
Equation (7) should be used to get the value for peak
vertical as well as peak horizontal dynamic pressure.
755
The value calculated from equation (7) is in addition to
static soil pressure.
Conclusion
A study of seismic response of a circular tunnel was
made using shake table testing and finite-element
numerical modelling. The compaction of sand was car-
ried out to maximum possible densities along the depth,
to avoid settlement during shaking. The bender element
test is used to get mechanical properties of materials.
The numerical results were compared with experimental
measurements in terms of peak dynamic stresses. The
comparison shows the numerical results are in good
agreement with the test results.
The civil engineering structures are designed against
the peak stresses and in case of flexible tunnel peak
horizontal stresses in soil near tunnel face are less than
the stresses observed at some distance from the tunnel.
Overall, there is a reduction of peak dynamic stresses in
soil as the flexibility of tunnel increases. In addition,
with increase in C/D ratio, the peak dynamic stresses
756 Measurement and Control 52(7-8)

Table 4. Seismic zone factor (z). 5. Lanzano G, Bilotta E, Russo G, et al. Experimental and
numerical study on circular tunnels under seismic load-
Seismic zone factor II III IV V ing. Eur J Environ Civ Eng 2015; 19: 539–563.
6. Chen J, Jiang L, Li J, et al. Numerical simulation of
Z 0.10 0.16 0.24 0.36 shaking table test on utility tunnel under non-uniform
earthquake excitation. Tunn Undergr Sp Technol 2012;
30: 205–216.
7. Chen J, Shi X and Li J. Shaking table test of utility tun-
Table 5. Design coefficient (K).
nel under non-uniform earthquake wave excitation. Soil
Dyn Earthq Eng 2010; 30: 1400–1416.
C/D Z-II Z-III Z-IV Z-V
8. Ling HI, Mohri Y, Kawabata T, et al. Centrifugal mod-
0.5–1 0.04 0.06 0.09 0.12 eling of seismic behavior of large-diameter pipe in liquefi-
1–3 0.07 0.11 0.16 0.18 able soil. J Geotech Geoenviron Eng 2003; 129: 1092–
3–5 0.11 0.17 0.21 0.22 1101.
9. Pitilakis D, Dietz M, Wood DM, et al. Numerical simu-
lation of dynamic soil–structure interaction in shaking
table testing. Soil Dyn Earthq Eng 2008; 28: 453–467.
decrease by significantly. Tunnel with smaller diameter 10. Cilingir U and Gopal Madabhushi SP. A model study on
generates larger peak dynamic stresses in soil less as the effects of input motion on the seismic behaviour of
compared to the larger diameter tunnel at the same C/ tunnels. Soil Dyn Earthq Eng 2011; 31: 452–462.
D ratio and flexibility ratio. Therefore, it is better to 11. Cilingir U and Madabhushi SPG. Effect of depth on seis-
have a flexible support to reduce the impact of such mic response of circular tunnels. Can Geotech J 2011; 48:
forces. It is also observed that, while designing the tun- 117–127.
12. Buonsanti M and Leonardi G. 3-D simulation of tunnel
nel, peak vertical stresses also need to be considered, as
structures under blast loading. Arch Civ Mech Eng 2013;
they are predominant at higher frequencies, especially 13: 128–134.
when the tunnel is placed in longitudinal direction of 13. Tsinidis G, Pitilakis K, Madabhushi G, et al. Dynamic
shaking. response of flexible square tunnels: centrifuge testing and
validation of existing design methodologies. Géotechnique
Acknowledgements 2015; 65: 401–417.
14. Huo H, Bobet A, Fernández G, et al. Load transfer
The authors wish to acknowledge the Department of Science mechanisms between underground structure and sur-
and Technology – SERB (grant no.: SB/S3/CEE/0002/2013) rounding ground: evaluation of the failure of the Daikai
for providing the financial support to carry out this research station. J Geotech Geoenviron Eng 2005; 131: 1522–1533.
work. 15. Penzien J. Seismically induced racking of tunnel linings.
Earthq Eng Struct Dyn 2000; 29: 683–691.
Declaration of conflicting interests 16. Lombardi D, Bhattacharya S, Scarpa F, et al. Dynamic
response of a geotechnical rigid model container with
The author(s) declared no potential conflicts of interest with absorbing boundaries. Soil Dyn Earthq Eng 2015; 69: 46–
respect to the research, authorship, and/or publication of this 56.
article. 17. ASTM D3574-17. Standard test methods for flexible cellu-
lar materials – slab, bonded, and molded urethane foams.
Funding West Conshohocken, PA: ASTM International.
18. Camacho-Tauta J, Álvarez-Jiménez JD and Reyes-Ortiz
The author(s) received no financial support for the research,
OJ. A procedure to calibrate and perform the bender ele-
authorship, and/or publication of this article.
ment test. Dyna 2012; 79: 10–18.
19. Kataoka S, Kawaguchi T and Horita T. Unconfined
ORCID iD compression strength and elastic shear modulus of air-
form treated lightweight soil. In: Proceedings of the 5th
DK Singh https://orcid.org/0000-0001-5618-9752
international symposium on deformation characteristics of
geomaterials, Seoul, South Korea, 1–3 September 2011,
pp. 682–686. Amsterdam: IOS Press.
References 20. Cilingir U. Seismic response of tunnels. Cambridge: Geo-
1. Yashiro K, Kojima Y and Shimizu M. Historical earth- technical Engineering Group, 2009.
quake damage to tunnels in Japan and case studies of 21. IS 1893 Part-1. Criteria for earthquake resistant design of
railway tunnels in the 2004 Niigataken-Chuetsu earth- structures, part-1 general provisions, and buildings. New
quake. Q Rep RTRI 2007; 48: 136–141. Delhi, India: Bureau of Indian Standard, 2016.
2. Wang J. Seismic design of tunnels: a simple state-of-the- 22. IS 4480 Part-5. Code of practice for design of tunnels con-
art design approach. New York: Parsons Brinckerhoff, veying water, part v structural design of concrete lining in
1993. soft strata and soils. New Delhi, India: Bureau of Indian
3. Clark C. Practice of dynamic pressure measurement. Standard, 2000.
Meas Control 1989; 22: 298–301. 23. ASCE/SEI 7-16. Minimum design loads for buildings and
4. White GWT. Methods of modelling and simulating other structures. Reston, VA: American Society of Civil
dynamic systems. Meas Control 1971; 4: T139–T146. Engineers, 1994.
Singh et al. 757

Appendix 1 R radius of the tunnel lining

t thickness of the tunnel lining
Notation vp p-wave velocity
A1 area in between loading-unloading curve vs s-wave velocity
A2 area under unloading curve n Poisson’s ratio
E Young’s modulus nm Poisson’s ratio of the medium
Es modulus of elasticity of the medium nt Poisson’s ratio of the tunnel
Et modulus of elasticity of the tunnel lining
G shear modulus r density of material
I moment of inertia of the tunnel lining (per
unit width)